Gary Marcus: AI cyberdefense introduces new attack vectors — just ask OpenAI
GaryMarcus · x · 2026-09-20
Gary Marcus cautions against David Sacks' claim that the best defense against cyberattacks is more AI. Citing Heidy Khlaaf's analysis, he notes that AI-powered cyber defense introduces unmitigated attack vectors that may defeat or worsen the advantages it was meant to provide — and "defensive AI isn't magic," pointing out OpenAI got hacked despite using it.
More from Safety
- Stab-a-figure test: Fable refuses all 20 trials while Astra complies in 85% — AaronBergman18 · 2026-09-20
- Helen Toner, who voted to oust Altman, on AI safety and overseeing OpenAI — trevposts · 2026-09-20
- repligate slams Anthropic-style eval logic: trained outputs aren't evidence — repligate · 2026-09-20
- LiveOverflow says OpenAI CISO apologized over vulnerability disclosure fumble — xeophon · 2026-09-20
- Paper Mill Studies Cited in 480 Patents and Clinical Guidelines, Analysis Finds — nicklaslundblad · 2026-09-20
- Lawsuit alleges Anthropic, OpenAI and others made illegal agreement to slow down AI — embedding-shape · 2026-09-20